A Survey of Encrypted Malicious Traffic Detection

Abstract

With more and more encrypted traffic such as HTTPS, encrypted traffic protects not only normal traffic, but also malicious traffic. Identification of encrypted malicious traffic without decryption has become a research hotspot. Combined with deep learning, an important branch of machine learning, encrypted malicious traffic detection has achieved good results. This paper reviews the detection of encrypted malicious traffic in recent years. Firstly, we classify encrypted malicious traffic. Secondly, we sorts out the extraction characteristics of encrypted malicious traffic, the key and difficult problems we are facing at present. Then, with encrypted malicious traffic detection technology as the main line, we summarized the current detection model from the four core aspects of data collection, data processing, model training and evaluation improvement. Finally, we analyze the problems and point out future research directions.